A5049311696- Statistical and numerical algorithms
- Scientific Measurement and Uncertainty Evaluation
- Control Systems and Identification
- Fire effects on ecosystems
- GNSS positioning and interference
- Advanced Frequency and Time Standards
- Diverse Scientific and Engineering Research
- Sparse and Compressive Sensing Techniques
- Ionosphere and magnetosphere dynamics
- Plant and animal studies
- Rangeland and Wildlife Management
- Botany, Ecology, and Taxonomy Studies
- Fire Detection and Safety Systems
- Fault Detection and Control Systems
- Radioactive Decay and Measurement Techniques
- Statistical Methods and Inference
- Remote Sensing in Agriculture
- Ecology and Vegetation Dynamics Studies
- Solar and Space Plasma Dynamics
- Landslides and related hazards
- Astronomy and Astrophysical Research
North Carolina State University
2023-2025
University of Michigan
2020
Abstract Astronomers often deal with data where the covariates and dependent variable are measured heteroscedastic non-Gaussian error. For instance, while TESS Kepler datasets provide a wealth of information, addressing challenges measurement errors systematic biases is critical for extracting reliable scientific insights improving machine learning models’ performance. Although techniques have been developed estimating regression parameters these data, few exist to construct prediction...
Abstract Ion upflow in the F region and topside ionosphere can greatly influence ion density fluxes at higher altitudes thus has significant impact on outflow. We investigated statistical characteristics of downflow using a 3‐year (2011–2013) data set from Poker Flat Incoherent Scatter Radar (PFISR). is twice more likely to occur nightside than dayside PFISR observations, while events often afternoon sector. Upflow tend ~500 km, those nightside. Both frequently as convection speed increases....
Abstract The aim of our paper is to investigate the properties classical phase-dispersion minimization (PDM), analysis variance (AOV), string-length (SL), and Lomb–Scargle (LS) power statistics from a statistician’s perspective. We confirm that when data are perturbations constant function, i.e. under null hypothesis no period in data, scaled version PDM statistic follows beta distribution, AOV an F LS chi-squared distribution with two degrees freedom. However, SL does not have closed-form...
Astronomers often deal with data where the covariates and dependent variable are measured heteroscedastic non-Gaussian error. For instance, while TESS Kepler datasets provide a wealth of information, addressing challenges measurement errors systematic biases is critical for extracting reliable scientific insights improving machine learning models' performance. Although techniques have been developed estimating regression parameters these data, few exist to construct prediction intervals...
The aim of our paper is to investigate the properties classical phase-dispersion minimization (PDM), analysis variance (AOV), string-length (SL), and Lomb-Scargle (LS) power statistics from a statistician's perspective. We confirm that when data are perturbations constant function, i.e. under null hypothesis no period in data, scaled version PDM statistic follows beta distribution, AOV an F LS chi-squared distribution with two degrees freedom. However, SL does not have closed-form...
Given the increasing prevalence of wildland fires in Western US, there is a critical need to develop tools understand and accurately predict burn severity. We machine learning model post-fire severity using pre-fire remotely sensed data. Hydrological, ecological, topographical variables collected from four regions California - sites Kincade fire (2019), CZU Lightning Complex (2020), Windy (2021), KNP Fire (2021) are used as predictors difference normalized ratio. hypothesize that Super...
The yaglm package aims to make the broader ecosystem of modern generalized linear models accessible data analysts and researchers. This encompasses a range loss functions (e.g. linear, logistic, quantile regression), constraints positive, isotonic) penalties. Beyond basic lasso/ridge, supports structured penalties such as nuclear norm well group, exclusive, fused, lasso. It also more accurate adaptive non-convex SCAD) versions these that often come with strong statistical guarantees at...